GitLab

more checking of the config file and its contents.

the frames and robots; always put the contents of the wiggle_frame
back into the pool, otherwise, stale frames can hang around and cause
problems (duh); clear the wiggling flag after we receive the idle
status update.

gatech works.

the default one.

tree, sticking it into @prefix@/src/emulab-src.tar.gz. This is to make
it slightly easier for other people to mess with elabinelab since at
the moment all elabinelabs are running from my private source tree!

will be stored in /usr/testbed/src. The node requesting it has to be
part of an elabinelab experiment, and of course has to provide the
appropriate secret key.

Also make a check to see if a client has a newer version than us and
complain if so.  Indicates that an older tmcd got installed.

Change one of the proctitle calls to not use info that is not yet available.

been moved to jibe with what the vision system reports as their
position.

Drop the backup table I create when updating the permissions table so
that schema-scheck does not complain.

and off.

There are two different kinds of admission control checks, hence two
locations.

* From tbswap, easy checks are made (like, too many nodes of a real
  type), or trying to the use robot lab when its closed.

* From assign_wrapper.in. Because of virtual types, and because assign
  is the ultimate descision maker, we need to pass into assign per
  physical type limits, and let it deal with how virtual types map to
  physical types, and whether it exceeds the limits. So, when called
  from assign_wrapper, libadminctrl returns a hash of types and the
  counts the current experiment is limited to. That is put into the
  ptop file where assign deals with it.

If admission control goes sour, then just turn off the sitevar for it:
'swap/use_admission_control'

admission control is turned off.

before the exec to avoid holding open the mysqld socket in the new
process.

table, but also that there are any nodes of that type. The entry could
have been put into the table because of a set-type-limit line in the
ptop file.

previously, since it has more information

vision system bounds).

number.

closed on Univ Holidays. I hope this daemon is gone before 2006
arrives!

Russ, I have no idea why this is here or how to get rid of it.

	my $out_width  = 792;
	my $out_height = 492;

Is there some way to get rid of these hardwired constants please?

Lbs

sitevars to determine if the Robot Lab is open or closed. The sitevars:

* 'robotlab/override' - Override other settings and forcibly turn the lab
  "on" or "off" (open or close). When the lab is turned off, new
  experiments cannot swap in and the current experiment is immediately
  swapped out.

* 'robotlab/exclusive' - The robot lab is exclusive use. Best to not mess
  with this sitevar :-)

* 'robotlab/opentime' - The time that the robot lab opens in the
  morning. The default is 07:00, but feel free to change this as you like.

* 'robotlab/closetime' - The time that the robot lab closes in the
  evening. The default is 18:00, but feel free to change this as you like.

* 'robotlab/open' - The robot lab is open or closed. DO NOT MESS WITH THIS!
  It is updated by the robomonitord script and intended to be used by
  admission control (not done yet).

The robomonitord script runs and periodically (every 2 minutes) wakes up
and looks at the various sitevars above. The lab is open during the day,
Monday through Friday, and closed on weekends. It is also supposed to be
closed on holidays, but I have not added that yet.

15 minutes before the lab is to be closed, a warning message is sent to the
swapper of the experiment running on the robot testbed, that their
experiment is going to be swapped soon. When the Robot lab is closed
(either cause the close time was reached, or because the lab was forcibly
closed with the override), the current experiment is immediately swapped
out.

I know, this is hopelessly bogus, but it will do until we feel like adding
a "Lab" datatype to the system.

in the other direction if it is on.